Primary Model Specification
Two-Way Fixed Effects Difference-in-Differences (TWFE DID) Approach Model Specification:
\[ \begin{aligned} \mathrm{wasting}_i &= \beta_0 + \beta_1 \cdot \mathrm{treated}_{it} + \sum_{j} \gamma_j \cdot \mathrm{time\_to\_treatment}_{it} \\ &\quad + \beta_2 \cdot \mathrm{child\_age}_{it} + \beta_3 \cdot \mathrm{child\_gender}_{it} + \beta_4 \cdot \mathrm{mother\_age}_{it} \\ &\quad + \beta_5 \cdot \mathrm{breastfed\_day}_{it} + \beta_6 \cdot \mathrm{ethnicity}_{it} + \beta_7 \cdot \mathrm{total\_income}_{it} \\ &\quad + \beta_8 \cdot \mathrm{region}_{it} + \beta_9 \cdot \mathrm{district}_{it} + \mu_i + \tau_t + \varepsilon_{it} \end{aligned} \]
Coefficients
- \(\mathrm{wasting}_{it}\): Dichotomous dependent variable indicating the status of wasting for child (i) at time (t).
- \(\mathrm{treated}_{it}\): Binary independent variable indicating whether child (i) received the treatment at time (t).
- \(\mathrm{time\_to\_treatment}_{it}\): Set of dummy variables representing the relative time to treatment, with (-1) as the reference category.
- \(\mathrm{child\_age}_{it}\): Control variable indicating age of child.
- \(\mathrm{child\_gender}_{it}\): Control variable indicating whether the child is a girl or a boy.
- \(\mathrm{mother\_age}_{it}\): Control variable indicating age of mother.
- \(\mathrm{breastfed\_day}_{it}\): Control variable indicating number of times the child is breastfed per day.
- \(\mathrm{ethnicity}_{it}\): Control variable indicating ethnicity of child.
- \(\mathrm{total\_income}_{it}\): Control variable indicating total household income.
- \(\mathrm{region}_{it}\): Control variable indicating region where child’s household is located.
- \(\mathrm{district}_{it}\): Control variable indicating district where child’s household is located.
- \(\mu_i\): Household fixed effects to control for unobserved household-level characteristics.
- \(\tau_t\): Year fixed effects to control for time-specific effects.
- \(\varepsilon_{it}\): Error term for unobserved factors affecting wasting.
Key Considerations
Individual Fixed Effects (FEs):
Estimation includes only individuals who change treatment status to avoid perfect collinearity between individual FEs and the treatment variable.Survey Wave Inclusion:
It is acceptable if individuals appear in the first wave of the survey or not, as this does not affect the initial estimation.
| Estimate | Std. Error | t value | Pr(>|t|) | Significance | |
|---|---|---|---|---|---|
| treated | 1.995 | 24339.793 | 0.000 | 1.000 | |
| time_to_treatment_adj::-2 | -0.129 | 0.100 | -1.283 | 0.199 | |
| time_to_treatment_adj::0 | 0.074 | 0.058 | 1.284 | 0.199 | |
| time_to_treatment_adj::1 | 0.052 | 0.066 | 0.789 | 0.430 | |
| time_to_treatment_adj::2 | 0.060 | 0.070 | 0.863 | 0.388 | |
| time_to_treatment_adj::3 | 0.158 | 0.085 | 1.847 | 0.065 | |
| child_age | -0.007 | 0.006 | -1.134 | 0.257 | |
| child_sex_cleanGirl | -0.005 | 0.011 | -0.411 | 0.681 | |
| breastfed_day_clean1-10 times | 0.015 | 0.015 | 1.010 | 0.312 | |
| breastfed_day_clean11-20 times | 0.071 | 0.020 | 3.492 | 0.000 | *** |
| breastfed_day_clean21 or more times | 0.043 | 0.029 | 1.512 | 0.131 | |
| ethnicity_cleanNewar | -0.078 | 1958.478 | 0.000 | 1.000 | |
| ethnicity_cleanTerai Brahmin/Chhettri | -0.047 | 23207.007 | 0.000 | 1.000 | |
| total_income_clean | 0.000 | 0.000 | 0.500 | 0.617 |